MMDetection

MMDetection

MMDetection is an open-source object detection toolbox developed by OpenMMLab. It provides a clean, unified framework for implementing and benchmarking detection algorithms, with support for two-stage detectors (Faster R-CNN), one-stage detectors (FCOS, ATSS), transformer-based models (DETR, DINO), and instance segmentation.

Key Features

• 40+ detection algorithms and 300+ pretrained models on COCO/VOC/Objects365
• Modular design: backbone → neck → head pipeline with drop-in replacements
• Distributed training (DDP) and mixed-precision (AMP) out of the box
• MMEngine training loop with metric logging, checkpointing, and LR scheduling
• New MMDet3D branch for 3D object detection and point cloud tasks

Quick Start

pip install mmdet
mim download mmdet --config rtmdet_tiny_8xb32-300e_coco --dest .

from mmdet.apis import init_detector, inference_detector

config = "rtmdet_tiny_8xb32-300e_coco.py"
checkpoint = "rtmdet_tiny_8xb32-300e_coco_20220902_112414-78e30dcc.pth"
model = init_detector(config, checkpoint, device="cuda:0")
result = inference_detector(model, "demo.jpg")

npx ai-supply add mmdetection-detection-framework

Curated mirror of the open-source MMDetection (Apache-2.0). Get it from the source.